Producing integrated circuit patterns on a wafer is technically demanding in two respects: first, integrated circuits contain a large amount of pattern information, with up to 2.times.10.sup.10 pattern features on a given process level of the chip. Second, the minimum feature size continues to shrink by a factor of two every six years, with 130 nm features expected in full scale manufacturing by the year 2003. These requirements place ever increasing demands on both the speed and spatial resolution of the lithographic process by which patterns are created on the chip.
Present day lithography is typically performed by first creating the pattern on a transparent reticle, then optically projecting this pattern onto the wafer using ultraviolet light. This projection process is repeated multiple times, once for each chip on the wafer. The spatial fidelity of the pattern on the chip is limited by the ability of the optical projection system to resolve the features, due to diffraction. This limitation, which occurs even in the case of a hypothetical perfect reticle, is severe with light optics. In addition, reticle patterns inevitably have imperfections, which also limit pattern fidelity on the wafer. Moreover, fabrication of the reticle is time consuming and expensive.
A well known solution is to use a focussed electron beam (e-beam) to write the pattern directly on the wafer. The diffraction limited resolution of an electron beam is, for this purpose, unlimited, owing to the fact that the wavelength of fast electrons is roughly a factor of 10.sup.4 smaller than ultraviolet light. In addition the need for a patterned reticle is eliminated, as the pattern is stored in computer memory, from which it is transferred directly to the wafer, without the need for projection.
Unfortunately, existing approaches to direct write e-beam, as it is called, are too slow to be practical for large scale manufacturing. The writing speed, measured in units of area swept out per unit time, is equal to the ratio of the total writing current to the dose measured in units of charge per unit area. The writing current needed to obtain minimally acceptable writing speed is of the order of 100 .mu.A. This corresponds to a writing speed of 10 cm.sup.2 /sec at a dose of 10 .mu.C/cm.sup.2.
Existing e-beam systems are not capable of delivering this minimum current. The reason is that, as the current is increased, the Coulomb interaction between beam electrons occurring in the beam path causes the image to blur, amounting to a loss of resolution. Coulomb interaction arises from the fact that the electrons are discrete charged particles, randomly distributed within the beam. The electrons thus exert a repulsive force on one another, which is random in magnitude and direction. The amount of blurring depends on the current density throughout the entire optical path. This in turn depends intimately on the optical design configuration of the system.
Systems with narrow beams illuminating a limited area or field, are called "probe-forming" systems. Because the beam is narrow, the electrons remain in close proximity to one another throughout the duration of their flight. The strength of the Coulomb interaction is inversely proportional to the square of the interelectron separation. In probe forming systems the separation is small, and the Coulomb interaction strong, thus limiting the usable writing current. The current in a typical probe forming system is on the order of 0.5 .mu.A, for which the blurring is 50 nm. This resolution is marginally acceptable for printing 180 nm wide lines, but the writing speed is at least a factor of 200 too slow.
A potential solution is to spread the writing current over a larger volume, thus reducing the Coulomb blurring. An example is electron projection printing, in which an electron optical image of a reticle can be used to write a large pattern area in a single flash. This is undesirable, however, for two reasons: first, it requires a patterned reticle. Second, all of the writing current is constrained to pass through a single, small aperture, which is needed to produce contrast in the image. Because of this constriction, electron projection suffers from Coulomb blurring as with other probe forming systems, though to a lesser extent.
In contrast to probe forming systems, a system in which the current is spread out over a large volume throughout the optical path may be termed "distributed". One way to construct such a system is by the use of multiple beams. The total writing current is given by the number of beams times the current in each beam. If the number of beams is large enough, the current in each beam may be made sufficiently low that Coulomb blurring does not impair the resolution.
The essential elements of such a system include, for each beam, an electron source, a lens, a means for positioning the beam relative to the writing surface, and a means for determining the numerical aperture (NA) of the optics. Each of these elements must be made to act identically on each beam, in a way that all beams act properly in concert, and maintain the correct relationship with one another.
Yasuda, et al. (U.S. Pat. No. 5,359,202) employ a single electron source, flooding an array of blanking apertures to produce a multiplicity of beams. The bundle of beams then passes through an optical system including a single aperture which defines the NA. As noted above, this constriction makes the system susceptible to Coulomb blurring, thereby limiting the useful current. Despite the use of multiple beams, this is not a distributed system, but a probe forming system.
J. E. Schneider, et al. (Journal of Vacuum Science and Technology, B 14(6), p. 3782 (1996)) employ a semiconductor on glass photocathode as the source for a parallel, multiple electron beam system for lithography. This system also constrains the writing current to pass through a single aperture which defines the NA. As the authors correctly point out, the useful current is limited by the Coulomb interaction to about 10 .mu.A. This is about a factor of 10 too small to be of practical use for manufacturing.
MacDonald (U.S. Pat. No. 5,363,021) employs a massively parallel array cathode consisting of a multiplicity of individual electron sources, explicitly, field emission tips. Each beam projects a single pixel onto the writing surface. This requires the source brightness to be relatively high, in order that the total writing current be suitable for high writing speed. The functions of focussing and positioning are accomplished by a separate lens and deflector for each beam. These structures consist of microscopically small elements precisely positioned relative to one another. Care must be taken to ensure precise alignment of each beam relative to its optics, as failure to do so would result in aberrations which degrade resolution. There is no mention of how the numerical aperture (NA) is determined. It is reasonable to assume, however, that each beam has a separate beam defining aperture which determines the NA.
A problem common to distributed systems is that of alignment. in the prior art, the well-known problems of aligning as system are multiplied by N, the number of modules in the system. The art has sought a distributed system that avoids alignment problems.